Prepare and test your skills

Prepare and test your skills

Active Assist insights are automated findings that help improve cost, performance, and reliability in Google Cloud. The Recommender API (opens in a new tab) analyzes usage and configuration data and produces a finding plus a suggested action. You review those results in the Cloud Assist panel in the Google Cloud console by asking Gemini Cloud Assist (opens in a new tab) in natural language.
Active Assist insights describe underused capacity, risky configuration, and efficiency opportunities that you can act on. Cost and efficiency insights point to idle virtual machines (VMs) and BigQuery datasets that should be partitioned or clustered, where you pay for capacity or scans you do not need. Reliability insights point to quotas that are almost at their limit and to unhealthy cluster configurations, where continued growth could block new VMs, jobs, or requests.
Common starting points for finding these insights are:
Cost findings save money without changing behavior, while reliability findings prevent outages when demand grows.
Access to Active Assist insights depends on an enabled API plus the correct Identity and Access Management (IAM) role. General resource recommendations need the Recommender API (recommender.googleapis.com) enabled in the project plus the Recommender Viewer role, and data then flows from the Recommender API to the Cloud Assist panel. Cost anomaly analysis has separate requirements and needs the App Optimize API (appoptimize.googleapis.com) plus the Cloud Hub Operator role.
Gemini Cloud Assist generally does not return direct billing amounts because many factors affect pricing. You visit the FinOps hub for the actual bill and detailed charge data.
Exam tip: General resource recommendations need the Recommender API plus Recommender Viewer role, while cost anomaly analysis needs the App Optimize API plus Cloud Hub Operator role.
Prioritizing Active Assist insights means comparing the risk of doing nothing against the benefit of acting. Quota and reliability findings have the highest operational impact because a quota at its limit can block autoscaling and new deployments. BigQuery partitioning and cluster health come next when slow queries or unstable workloads affect performance. Idle VM findings have the clearest cost benefit because stopping or downsizing removes spend with low risk.
A decision tree that prioritizes Active Assist insights from quota limits and cluster reliability first, then BigQuery performance, to idle VM cost savings last.
| Insight | What it flags | Prioritize when |
|---|---|---|
| Quota near limit | Usage close to project quota | Growth or peak traffic could soon be blocked |
| Cluster reliability | Risky or unhealthy cluster configuration | Availability or workload stability is affected |
| BigQuery partitioning and clustering | Large tables scanned inefficiently | Query speed and scan cost matter for the workload |
| Idle VMs and cost reduction | Low-utilization resources with estimated savings | You want quick savings with low operational risk |
Apply fixes that prevent outages first, then fixes that improve performance, then fixes that only save money.
Implement Automation and Governance Using Active Assist means combining intelligent recommendations with automatic guardrails to keep resource use efficient across many projects. Active Assist is a portfolio of intelligent tools that analyzes Google Cloud usage and suggests actions to reduce cost, improve performance, and support sustainability. Automation and governance then apply those ideas consistently, so teams do not have to check every resource by hand. Together, recommendations plus controls enforce efficient practices across projects and environments.
Active Assist provides targeted suggestions such as VM rightsizing for Compute Engine managed instance groups, which matches machine size to actual demand. The FinOps hub extends this idea for spending and uses an AI engine to recommend savings for resources that are currently deployed, with access limited by role-based access control (RBAC). Teams that foster a culture of cost awareness (opens in a new tab) use these recommendations to continuously review spend by project, department, or cost center.
Cloud Monitoring complements recommendations by tracking CPU, memory, trends, and alerts for critical resources. Utilization data flows from Cloud Monitoring into decisions to downsize or reconfigure underused resources. This link between measured use and recommended action helps teams optimize resource usage (opens in a new tab) instead of guessing.
Reducing manual oversight with automation means letting alerts and built-in adjustments handle routine optimization work. Automated notifications inform stakeholders through email or chat about cost anomalies, budget thresholds, and new savings opportunities. Proactive quota alerts warn when project-level quota use reaches a percentage of the maximum, so teams can request increases before deployments fail.
Automation also applies standards and sizing without a manual step for each resource. Automation tools like Terraform can enforce labeling on every created resource, which improves cost visibility and attribution. Built-in product features adjust capacity to demand automatically, such as autoscaling for Compute Engine and Google Kubernetes Engine (GKE) and Object Lifecycle Management and Autoclass for Cloud Storage.
Organization-level governance enforces consistent optimization using policies, quotas, and budgets that apply across projects and environments. Organization policies can limit which Google Cloud locations and products teams may use, so resources are provisioned where they align with cost and compliance goals. Detailed budgets for each project, workload, and environment set spending thresholds that are monitored against actual use and reviewed as needs change.
Quotas provide granular control over how much can be consumed at different levels:
Quotas and budgets together create layered cost control, where quotas block excess provisioning and budgets warn about excess spend.
Exam tip: Active Assist provides recommendations such as rightsizing, while organization policies, quotas, and budgets enforce limits on what can be provisioned.
Apply Recommendation Hub and Product-Specific Advisors means using one central view plus specialized advisors to act on the right optimization at the right time. Active Assist is the collection of Google Cloud tools that uses machine learning to improve cost, performance, reliability, and security. The Recommendation Hub is the central place in the console where findings from product-specific advisors are collected, and the Recommender API (opens in a new tab) is the service that generates them. Each advisor produces an insight that describes what was observed and a recommendation that describes what to do.
The Recommendation Hub gives a single view of guidance from many advisors instead of checking each service separately. Access depends on the Recommender API being enabled and on holding the Recommender Viewer IAM role, which controls who can view recommendations. Data flows from the Recommender API to the console recommendation panel, where you select a recommendation to see detailed next steps, or to Gemini Cloud Assist (opens in a new tab) when you ask in chat.
Common requests that return Hub findings are:
The console path suits browsing many recommendations, while the chat path suits asking a focused question about one project.
Exam tip: An insight describes what was observed, such as unused permissions, while a recommendation describes what to do, such as remove a role.
The Identity and Access Management (IAM) Recommender (opens in a new tab) is the product-specific advisor for access control and is part of the Policy Intelligence suite. The IAM Recommender uses machine learning to identify accounts and permissions that are no longer needed, such as when an employee changes roles or a service account was abandoned after a task finished. Applying its guidance supports least privilege, which means each member keeps only the permissions actually used.
A common result has the subtype REMOVE_ROLE, which signals that a granted role appears unused and can be removed. Related insights highlight permission usage patterns and unused service accounts across a project. For review at scale, recommendation data can be exported to BigQuery with the BigQuery Data Transfer Service and then queried and visualized with tools such as Looker, so administrators can rightsize access across many projects before applying changes.
Compute and data advisors focus on waste, query shape, and capacity risk rather than access. An idle-VM finding signals a Compute Engine instance that is running but not doing useful work, so stopping, downsizing, or deleting it reduces cost without affecting workloads. A BigQuery finding signals datasets that should be partitioned or clustered, so queries scan less data and run faster at lower cost. A quota finding signals limits that are almost reached, so you can request an increase before deployments fail or autoscaling is blocked.
Choose to apply a recommendation when the insight matches current need and the tradeoff is acceptable. Stop or rightsize an idle VM when its workloads have ended or moved, and keep it when it is standby capacity for failover or scheduled jobs. Partition or cluster a BigQuery dataset when large tables are scanned repeatedly, and act on a quota warning when growth or peak traffic will soon require more capacity.
An Active Assist insight describes what was observed, while a recommendation describes what action to take. For example, an insight may describe unused permissions and the paired recommendation suggests removing the role.
General resource recommendations require the Recommender API enabled in the project plus the Recommender Viewer role, with data flowing from the Recommender API to the Cloud Assist panel. Cost anomaly analysis separately requires the App Optimize API plus the Cloud Hub Operator role.
Quota findings near their limit have the highest priority because they can block new deployments, jobs, or autoscaling when demand grows. Idle virtual machines findings have the clearest cost benefit because stopping or downsizing removes spend with low risk, so apply fixes that prevent outages first and fixes that only save money after.
Organization policies, quotas, and budgets enforce consistent limits on what can be provisioned across projects and environments. Active Assist provides recommendations such as rightsizing that adjust usage to actual demand instead of blocking provisioning.
A media company runs several analytics and web workloads on Google Cloud Compute Engine virtual machines (VMs). A recent internal review indicates that compute infrastructure spending is significantly higher than necessary. Utilization data collected over the past month shows that several VMs consistently utilize less than 10% of their allocated CPU and RAM. The operations team wants an automated Active Assist tool that analyzes these utilization patterns and suggests specific, smaller instance configurations to optimize spending without degrading performance.
Which Active Assist tool or recommender should the operations team use?
Last updated